📄 中文摘要
研究将劳动经济学中的常规与非常规任务区分应用于 AI 自动化知识工作的可行性，探讨了在何种情况下 AI 自动化是结构上可行的，以及波兰尼悖论的适用性。提出了“可辩护但不具差异化”的认知劳动概念，认为这是加速 AI 应用的主要领域。该研究强调了在知识工作中，某些任务因其可预测性和重复性，更容易被 AI 技术所替代，而其他任务则因其复杂性和人类直觉的不可替代性，仍然需要人类的参与。
决定 AI 能够自动化哪些知识工作的因素
出处: What Determines Which Knowledge Work AI Can Actually Automate
发布: 2026年3月18日
📄 English Summary
What Determines Which Knowledge Work AI Can Actually Automate
The study applies the routine/non-routine task distinction from labor economics to assess the structural feasibility of AI automation in knowledge work. It explores the conditions under which AI automation is tractable and where Polanyi's Paradox applies. The concept of 'defensible but not differentiating' cognitive labor is introduced as a prime area for acceleration. The research emphasizes that certain tasks in knowledge work, due to their predictability and repetitiveness, are more likely to be replaced by AI technology, while others, due to their complexity and the irreplaceability of human intuition, will still require human involvement.